In [1]:
import numpy as np
import cv2, glob
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
from moviepy.editor import VideoFileClip
from IPython.display import HTML
from collections import deque
%matplotlib inline
In [2]:
class Error(Exception):
pass
In [3]:
#Camera Calibration
def camera_Calibraton(directory, filename, nx, ny, img_size):
objp = np.zeros((nx*ny,3), np.float32)
objp[:,:2] = np.mgrid[0:nx, 0:ny].T.reshape(-1,2)
objpoints = []
imgpoints = []
# Image List
images = glob.glob('/'+directory+'/'+filename+'*'+'.jpg')
# Step through the list and search for chessboard corners
for idx, fname in enumerate(images):
img = cv2.imread(fname)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# Find the chessboard corners
ret, corners = cv2.findChessboardCorners(gray, (nx, ny), None)
#print("name:",fname,"RET:",ret)
# If found, add object points, image points
if ret == True:
objpoints.append(objp)
imgpoints.append(corners)
if (len(objpoints) == 0 or len(imgpoints) == 0):
raise Error("Calibration Failed")
ret, mtx, dist, rvecs, tvecs = cv2.calibrateCamera(objpoints, imgpoints, img_size, None, None)
return mtx, dist
In [4]:
#Image undistort
def undistort(image, mtx, dist):
image = cv2.undistort(image, mtx, dist, None, mtx)
return image
In [5]:
#Perspective transform
def transform(undist,src,dst,img_size):
M = cv2.getPerspectiveTransform(src, dst)
Minv = cv2.getPerspectiveTransform(dst, src)
warped = cv2.warpPerspective(undist, M, img_size)
return warped, M, Minv
In [6]:
nx = 9
ny = 6
mtx, dist = camera_Calibraton('Users/rickerish_nah/Documents/trials/udacity Q/camera_cal', 'calibration', nx, ny, (720, 1280))
images_calib = glob.glob('/Users/rickerish_nah/Documents/trials/udacity Q/camera_cal/calibration**.jpg')
for i, img in enumerate(images_calib):
checker_dist = cv2.imread(img)
checker_dist = cv2.cvtColor(checker_dist, cv2.COLOR_BGR2RGB)
#Undistort
checker_undist = undistort(checker_dist, mtx, dist)
#Warp
gray = cv2.cvtColor(checker_undist, cv2.COLOR_BGR2GRAY)
ret, corners = cv2.findChessboardCorners(gray, (nx,ny), None)
img_size = (gray.shape[1], gray.shape[0])
if ret == True:
src = np.float32([corners[0], corners[nx-1], corners[-1], corners[-nx]])
offset = 100
dstn = np.float32([[offset, offset], [img_size[0]-offset, offset], [img_size[0]-offset, img_size[1]-offset], [offset, img_size[1]-offset]])
checker_warped, M, Minv = transform(checker_undist,src,dstn,img_size)
f, ((ax1, ax2,ax3)) = plt.subplots(1, 3, figsize=(12, 18))
ax1.imshow(checker_dist)
ax1.set_title('Original', fontsize=15)
ax2.imshow(checker_undist)
ax2.set_title('Undistorted', fontsize=15)
ax3.imshow(checker_warped)
ax3.set_title('Warped', fontsize=15)
In [7]:
def abs_sobel_thresh(image, orient, sobel_kernel, thresh):
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
assert(orient == 'x' or orient == 'y'), "Orientation must be x or y"
if orient == 'x':
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0,ksize = sobel_kernel)
abs_sobelx = np.absolute(sobelx)
scaled_sobel = np.uint8(255*abs_sobelx/np.max(abs_sobelx))
else:
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1,ksize = sobel_kernel)
abs_sobely = np.absolute(sobely)
scaled_sobel = np.uint8(255*abs_sobely/np.max(abs_sobely))
grad_binary = np.zeros_like(scaled_sobel)
grad_binary[(scaled_sobel >= thresh[0]) & (scaled_sobel <= thresh[1])] = 1
return grad_binary
def mag_thresh(image, sobel_kernel, thresh):
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize = sobel_kernel)
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize = sobel_kernel)
abs_sobelxy = np.power((np.power(sobelx,2)+np.power(sobely,2)),0.5)
scaled_sobel = np.uint8(255*abs_sobelxy/np.max(abs_sobelxy))
mag_binary = np.zeros_like(scaled_sobel)
mag_binary[(scaled_sobel >= thresh[0]) & (scaled_sobel <= thresh[1])] = 1
return mag_binary
def dir_threshold(image, sobel_kernel, thresh):
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel)
with np.errstate(divide='ignore', invalid='ignore'):
absgraddir = np.absolute(np.arctan(sobely/sobelx))
dir_binary = np.zeros_like(absgraddir)
dir_binary[(absgraddir >= thresh[0]) & (absgraddir <= thresh[1])] = 1
return dir_binary
In [8]:
def get_thresholded_image(img):
# apply gradient threshold on the horizontal gradient
sx_binary = abs_sobel_thresh(img, 'x', 3, thresh=(10, 200))
# apply gradient direction threshold so that only edges closer to vertical are detected.
dir_binary = dir_threshold(img, 3,thresh=(np.pi/6, np.pi/2))
# combine the gradient and direction thresholds.
combined_condition = ((sx_binary == 1) & (dir_binary == 1))
# R & G thresholds so that yellow lanes are detected well.
color_threshold = 150
R = img[:,:,0]
G = img[:,:,1]
color_combined = np.zeros_like(R)
r_g_condition = (R > color_threshold) & (G > color_threshold)
# color channel thresholds
hls = cv2.cvtColor(img, cv2.COLOR_RGB2HLS)
S = hls[:,:,2]
L = hls[:,:,1]
# S channel performs well for detecting bright yellow and white lanes
s_thresh = (100, 255)
s_condition = (S > s_thresh[0]) & (S <= s_thresh[1])
# We put a threshold on the L channel to avoid pixels which have shadows and as a result darker.
l_thresh = (120, 255)
l_condition = (L > l_thresh[0]) & (L <= l_thresh[1])
# combine all the thresholds
# A pixel should either be a yellowish or whiteish
# And it should also have a gradient, as per our thresholds
color_combined[(r_g_condition & l_condition) & (s_condition | combined_condition)] = 1
# apply the region of interest mask
mask = np.zeros_like(color_combined)
region_of_interest_vertices = np.array([[0,img.shape[0]-1], [img.shape[1]/2, int(0.5*img.shape[0])], [img.shape[1]-1, img.shape[0]-1]], dtype=np.int32)
cv2.fillPoly(mask, [region_of_interest_vertices], 1)
thresholded = cv2.bitwise_and(color_combined, mask)
return thresholded
In [9]:
images_test = glob.glob('/Users/rickerish_nah/Documents/trials/udacity Q/test_images/*.jpg')
for i, img in enumerate(images_test):
img = cv2.imread(img)#'/Users/rickerish_nah/Documents/trials/udacity Q/test_images/straight_lines1.jpg')
test_img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
test_img_dst = undistort(test_img,mtx, dist)
thresholded = get_thresholded_image(test_img)
# Vertices extracted manually for performing a perspective transform
bottom_left = [260,680]
bottom_right = [1050, 680]
top_left = [570, 470]
top_right = [725, 470]
source = np.float32([bottom_left,bottom_right,top_right,top_left])
pts = np.array([bottom_left,bottom_right,top_right,top_left], np.int32)
pts = pts.reshape((-1,1,2))
copy = img.copy()
cv2.polylines(copy,[pts],True,(255,0,0), thickness=3)
bottom_left = [200,680]
bottom_right = [1000, 680]
top_left = [200, 0]
top_right = [1000, 0]
dst = np.float32([bottom_left,bottom_right,top_right,top_left])
img_size = (test_img.shape[1],test_img.shape[0])
test_warp, M, Minv = transform(test_img_dst,source,dst,img_size)
thresh_warp, M, Minv = transform(thresholded,source,dst,img_size)
f, (ax1,ax2,ax3,ax4,ax5) = plt.subplots(1, 5, figsize=(20,30))
ax1.imshow(test_img)
ax1.set_title('Original Image', fontsize=8)
ax2.imshow(copy)
ax2.set_title('Undistorted Image', fontsize=8)
ax3.imshow(thresholded,cmap = 'gray')
ax3.set_title('Thresholded', fontsize=8)
ax4.imshow(test_warp)
ax4.set_title('Undistorted & Warped Image', fontsize=8)
ax5.imshow(thresh_warp, cmap = 'gray')
ax5.set_title('thresholded&warped', fontsize=8)
In [10]:
#Histogram to locate the lane line
def histogram(img_warped):
hist = np.sum(img_warped[img_warped.shape[0]//2:,:], axis=0)
# Peak in the first half indicates the likely position of the left lane
half_width = np.int(hist.shape[0]/2)
leftx_base = np.argmax(hist[:half_width])
# Peak in the second half indicates the likely position of the right lane
rightx_base = np.argmax(hist[half_width:]) + half_width
return hist, leftx_base, rightx_base
hist, leftx_base, rightx_base = histogram(thresh_warp)
print(leftx_base, rightx_base)
plt.plot(hist)
Out[10]:
In [13]:
#Sliding Window Technique
def find_lane_pixels(binary_warped):
out_img = np.dstack((binary_warped, binary_warped, binary_warped))
# HYPERPARAMETERS
# Choose the number of sliding windows
nwindows = 9
# Set the width of the windows +/- margin
margin = 60
# Set minimum number of pixels found to recenter window
minpix = 50
# Set height of windows - based on nwindows above and image shape
window_height = np.int(binary_warped.shape[0]//nwindows)
# Identify the x and y positions of all nonzero pixels in the image
nonzero = binary_warped.nonzero()
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
# Current positions to be updated later for each window in nwindows
leftx_current = leftx_base
rightx_current = rightx_base
# Create empty lists to receive left and right lane pixel indices
left_lane_inds = []
right_lane_inds = []
# Step through the windows one by one
for window in range(nwindows):
# Identify window boundaries in x and y (and right and left)
win_y_low = binary_warped.shape[0] - (window+1)*window_height
win_y_high = binary_warped.shape[0] - window*window_height
win_xleft_low = leftx_current - margin
win_xleft_high = leftx_current + margin
win_xright_low = rightx_current - margin
win_xright_high = rightx_current + margin
# Draw the windows on the visualization image
cv2.rectangle(out_img,(win_xleft_low,win_y_low),(win_xleft_high,win_y_high),(0,255,0), 3)
cv2.rectangle(out_img,(win_xright_low,win_y_low),(win_xright_high,win_y_high),(0,255,0), 3)
# Identify the nonzero pixels in x and y within the window #
good_left_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) &
(nonzerox >= win_xleft_low) & (nonzerox < win_xleft_high)).nonzero()[0]
good_right_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) &
(nonzerox >= win_xright_low) & (nonzerox < win_xright_high)).nonzero()[0]
# Append these indices to the lists
left_lane_inds.append(good_left_inds)
right_lane_inds.append(good_right_inds)
# If you found > minpix pixels, recenter next window on their mean position
if len(good_left_inds) > minpix:
leftx_current = np.int(np.mean(nonzerox[good_left_inds]))
if len(good_right_inds) > minpix:
rightx_current = np.int(np.mean(nonzerox[good_right_inds]))
# Concatenate the arrays of indices (previously was a list of lists of pixels)
try:
left_lane_inds = np.concatenate(left_lane_inds)
right_lane_inds = np.concatenate(right_lane_inds)
except ValueError:
# Avoids an error if the above is not implemented fully
pass
# Extract left and right line pixel positions
leftx = nonzerox[left_lane_inds]
lefty = nonzeroy[left_lane_inds]
rightx = nonzerox[right_lane_inds]
righty = nonzeroy[right_lane_inds]
return leftx, lefty, rightx, righty, out_img
def fit_polynomial(binary_warped):
# Find our lane pixels first
leftx, lefty, rightx, righty, out_img = find_lane_pixels(binary_warped)
# Fit a second order polynomial to each using `np.polyfit`
left_fit = np.polyfit(lefty, leftx, 2)
right_fit = np.polyfit(righty, rightx, 2)
# Generate x and y values for plotting
ploty = np.linspace(0, binary_warped.shape[0]-1, binary_warped.shape[0] )
try:
left_fitx = left_fit[0]*ploty**2 + left_fit[1]*ploty + left_fit[2]
right_fitx = right_fit[0]*ploty**2 + right_fit[1]*ploty + right_fit[2]
except TypeError:
# Avoids an error if `left` and `right_fit` are still none or incorrect
print('The function failed to fit a line!')
left_fitx = 1*ploty**2 + 1*ploty
right_fitx = 1*ploty**2 + 1*ploty
## Visualization ##
# Colors in the left and right lane regions
out_img[lefty, leftx] = [255, 0, 0]
out_img[righty, rightx] = [0, 0, 255]
bin_img=np.zeros_like(binary_warped)
bin_img[lefty, leftx] = 255
bin_img[righty, rightx] = 255
return out_img, left_fit, right_fit, ploty
out_img, left_fit, right_fit, ploty = fit_polynomial(thresh_warp)
left_fitx = left_fit[0]*ploty**2 + left_fit[1]*ploty + left_fit[2]
right_fitx = right_fit[0]*ploty**2 + right_fit[1]*ploty + right_fit[2]
plt.plot(left_fitx, ploty, color='yellow')
plt.plot(right_fitx, ploty, color='yellow')
plt.imshow(out_img)
Out[13]:
In [14]:
#Search from Prior
def search_around_poly(binary_warped,left_fit, right_fit):
# HYPERPARAMETER
# Choose the width of the margin around the previous polynomial to search
# The quiz grader expects 100 here, but feel free to tune on your own!
margin = 100
# Grab activated pixels
nonzero = binary_warped.nonzero()
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
### Set the area of search based on activated x-values ###
### within the +/- margin of our polynomial function ###
### Hint: consider the window areas for the similarly named variables ###
### in the previous quiz, but change the windows to our new search area ###
left_lane_inds = ((nonzerox > (left_fit[0]*(nonzeroy**2) + left_fit[1]*nonzeroy +
left_fit[2] - margin)) & (nonzerox < (left_fit[0]*(nonzeroy**2) +
left_fit[1]*nonzeroy + left_fit[2] + margin)))
right_lane_inds = ((nonzerox > (right_fit[0]*(nonzeroy**2) + right_fit[1]*nonzeroy +
right_fit[2] - margin)) & (nonzerox < (right_fit[0]*(nonzeroy**2) +
right_fit[1]*nonzeroy + right_fit[2] + margin)))
# Again, extract left and right line pixel positions
leftx = nonzerox[left_lane_inds]
lefty = nonzeroy[left_lane_inds]
rightx = nonzerox[right_lane_inds]
righty = nonzeroy[right_lane_inds]
# Fit new polynomials
ploty = np.linspace(0, binary_warped.shape[0]-1, binary_warped.shape[0])
if len(leftx)==0:
left_fit=[]
else:
left_fit = np.polyfit(lefty, leftx, 2)
if len(rightx)==0:
right_fit=[]
else:
right_fit = np.polyfit(righty, rightx, 2)
left_fitx = left_fit[0]*ploty**2 + left_fit[1]*ploty + left_fit[2]
right_fitx = right_fit[0]*ploty**2 + right_fit[1]*ploty + right_fit[2]
## Visualization ##
# Create an image to draw on and an image to show the selection window
out_img = np.dstack((binary_warped, binary_warped, binary_warped))*255
window_img = np.zeros_like(out_img)
# Color in left and right line pixels
out_img[nonzeroy[left_lane_inds], nonzerox[left_lane_inds]] = [255, 0, 0]
out_img[nonzeroy[right_lane_inds], nonzerox[right_lane_inds]] = [0, 0, 255]
# Generate a polygon to illustrate the search window area
# And recast the x and y points into usable format for cv2.fillPoly()
left_line_window1 = np.array([np.transpose(np.vstack([left_fitx-margin, ploty]))])
left_line_window2 = np.array([np.flipud(np.transpose(np.vstack([left_fitx+margin,
ploty])))])
left_line_pts = np.hstack((left_line_window1, left_line_window2))
right_line_window1 = np.array([np.transpose(np.vstack([right_fitx-margin, ploty]))])
right_line_window2 = np.array([np.flipud(np.transpose(np.vstack([right_fitx+margin,
ploty])))])
right_line_pts = np.hstack((right_line_window1, right_line_window2))
# Draw the lane onto the warped blank image
cv2.fillPoly(window_img, np.int_([left_line_pts]), (0,255, 0))
cv2.fillPoly(window_img, np.int_([right_line_pts]), (0,255, 0))
result = cv2.addWeighted(out_img, 1, window_img, 0.3, 0)
return result, left_fit, right_fit, ploty
# Run image through the pipeline
# Note that in your project, you'll also want to feed in the previous fits
out_img_prior, left_fit_new, right_fit_new, ploty = search_around_poly(thresh_warp,left_fit, right_fit)
left_fitx = left_fit_new[0]*ploty**2 + left_fit_new[1]*ploty + left_fit_new[2]
right_fitx = right_fit_new[0]*ploty**2 + right_fit_new[1]*ploty + right_fit_new[2]
# View your output
plt.plot(left_fitx, ploty, color='yellow')#_______________________
plt.plot(right_fitx, ploty, color='yellow')
plt.imshow(out_img_prior)
Out[14]:
In [15]:
#Measuring Radius of Curvature:
num_rows = test_warp.shape[0]
def measure_radius_of_curvature(x_values):
ym_per_pix = 30/720 # meters per pixel in y dimension
xm_per_pix = 3.7/700 # meters per pixel in x dimension
# If no pixels were found return None
y_points = np.linspace(0, num_rows-1, num_rows)
y_eval = np.max(y_points)
# Fit new polynomials to x,y in world space
fit_cr = np.polyfit(y_points*ym_per_pix, x_values*xm_per_pix, 2)
curverad = ((1 + (2*fit_cr[0]*y_eval*ym_per_pix + fit_cr[1])**2)**1.5) / np.absolute(2*fit_cr[0])
return curverad
left_curve_rad = measure_radius_of_curvature(left_fitx)
right_curve_rad = measure_radius_of_curvature(right_fitx)
average_curve_rad = (left_curve_rad + right_curve_rad)/2
curvature_string = "Radius of curvature: %.2f m" % average_curve_rad
print(curvature_string)
In [16]:
# compute the offset from the center
lane_center = (right_fitx[719] + left_fitx[719])/2
xm_per_pix = 3.7/700 # meters per pixel in x dimension
center_offset_pixels = abs(img_size[0]/2 - lane_center)
center_offset_mtrs = xm_per_pix*center_offset_pixels
offset_string = "Center offset: %.2f m" % center_offset_mtrs
print(offset_string)
In [17]:
#Inverse Perspective Transform for sliding window
for i, img in enumerate(images_test):
img = cv2.imread(img)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
test_img = get_thresholded_image(img)
img_size = (test_img.shape[1], test_img.shape[0])
test_warp, M, Minv = transform(test_img,source,dst, img_size)
out_img_prior, left_fit_new, right_fit_new, ploty = fit_polynomial(test_warp)#search_around_poly(test_warp,left_fit, right_fit)
left_fitx = left_fit_new[0]*ploty**2 + left_fit_new[1]*ploty + left_fit_new[2]
right_fitx = right_fit_new[0]*ploty**2 + right_fit_new[1]*ploty + right_fit_new[2]
#ROC
left_curve_rad = measure_radius_of_curvature(left_fitx)
right_curve_rad = measure_radius_of_curvature(right_fitx)
average_curve_rad = (left_curve_rad + right_curve_rad)/2
curvature_string = "Radius of curvature: %.2f m" % average_curve_rad
#Offset
lane_center = (right_fitx[719] + left_fitx[719])/2
xm_per_pix = 3.7/700 # meters per pixel in x dimension
center_offset_pixels = abs(img_size[0]/2 - lane_center)
center_offset_mtrs = xm_per_pix*center_offset_pixels
offset_string = "Center offset: %.2f m" % center_offset_mtrs
out_img = np.dstack((test_warp, test_warp, test_warp))*255
left_line_window = np.array(np.transpose(np.vstack([left_fitx, ploty])))
right_line_window = np.array(np.flipud(np.transpose(np.vstack([right_fitx, ploty]))))
line_points = np.vstack((left_line_window, right_line_window))
cv2.fillPoly(out_img, np.int_([line_points]), [0,255, 0])
unwarped = cv2.warpPerspective(out_img, Minv, img_size , flags=cv2.INTER_LINEAR)
result = cv2.addWeighted(img, 1, unwarped, 0.3, 0)
result = cv2.putText(result,curvature_string,(100, 90), cv2.FONT_HERSHEY_COMPLEX, 1.4, (255,255,255),2,cv2.LINE_AA)
result = cv2.putText(result,offset_string,(100, 150), cv2.FONT_HERSHEY_COMPLEX, 1.4, (255,255,255),2,cv2.LINE_AA)
f, (ax1, ax2) = plt.subplots(1, 2, figsize=(20,10))
ax1.imshow(img)
ax1.set_title('Original Image', fontsize=30)
ax2.imshow(result)
ax2.set_title('color and gradient thresholded: sliding window', fontsize=15)
In [18]:
#Inverse Perspective Transform: Priori
for i, img in enumerate(images_test):
img = cv2.imread(img)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
test_img = get_thresholded_image(img)
img_size = (test_img.shape[1], test_img.shape[0])
test_warp, M, Minv = transform(test_img,source,dst, img_size)
out_img_prior, left_fit_new, right_fit_new, ploty = search_around_poly(test_warp,left_fit, right_fit)
left_fitx = left_fit_new[0]*ploty**2 + left_fit_new[1]*ploty + left_fit_new[2]
right_fitx = right_fit_new[0]*ploty**2 + right_fit_new[1]*ploty + right_fit_new[2]
#ROC
left_curve_rad = measure_radius_of_curvature(left_fitx)
right_curve_rad = measure_radius_of_curvature(right_fitx)
average_curve_rad = (left_curve_rad + right_curve_rad)/2
curvature_string = "Radius of curvature: %.2f m" % average_curve_rad
#Offset
lane_center = (right_fitx[719] + left_fitx[719])/2
xm_per_pix = 3.7/700 # meters per pixel in x dimension
center_offset_pixels = abs(img_size[0]/2 - lane_center)
center_offset_mtrs = xm_per_pix*center_offset_pixels
offset_string = "Center offset: %.2f m" % center_offset_mtrs
out_img = np.dstack((test_warp, test_warp, test_warp))*255
left_line_window = np.array(np.transpose(np.vstack([left_fitx, ploty])))
right_line_window = np.array(np.flipud(np.transpose(np.vstack([right_fitx, ploty]))))
line_points = np.vstack((left_line_window, right_line_window))
cv2.fillPoly(out_img, np.int_([line_points]), [0,255, 0])
unwarped = cv2.warpPerspective(out_img, Minv, img_size , flags=cv2.INTER_LINEAR)
result = cv2.addWeighted(img, 1, unwarped, 0.3, 0)
result = cv2.putText(result,curvature_string,(100, 90), cv2.FONT_HERSHEY_COMPLEX, 1.4, (255,255,255),2,cv2.LINE_AA)
result = cv2.putText(result,offset_string,(100, 150), cv2.FONT_HERSHEY_COMPLEX, 1.4, (255,255,255),2,cv2.LINE_AA)
f, (ax1, ax2) = plt.subplots(1, 2, figsize=(20,10))
ax1.imshow(img)
ax1.set_title('Original Image', fontsize=30)
ax2.imshow(result)
ax2.set_title('color and gradient thresholded P', fontsize=30)
In [19]:
print("PROJECT 4: Advanced Lane Finding........")
In [20]:
import numpy as np
import cv2, glob
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
from moviepy.editor import VideoFileClip
from IPython.display import HTML
from collections import deque
%matplotlib inline
In [21]:
class Error(Exception):
pass
In [22]:
#Camera Calibration
def camera_Calibraton(directory, filename, nx, ny, img_size):
objp = np.zeros((nx*ny,3), np.float32)
objp[:,:2] = np.mgrid[0:nx, 0:ny].T.reshape(-1,2)
objpoints = []
imgpoints = []
# Image List
images = glob.glob('/'+directory+'/'+filename+'*'+'.jpg')
# Step through the list and search for chessboard corners
for idx, fname in enumerate(images):
img = cv2.imread(fname)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# Find the chessboard corners
ret, corners = cv2.findChessboardCorners(gray, (nx, ny), None)
#print("name:",fname,"RET:",ret)
# If found, add object points, image points
if ret == True:
objpoints.append(objp)
imgpoints.append(corners)
if (len(objpoints) == 0 or len(imgpoints) == 0):
raise Error("Calibration Failed")
ret, mtx, dist, rvecs, tvecs = cv2.calibrateCamera(objpoints, imgpoints, img_size, None, None)
return mtx, dist
In [23]:
#Image undistort
def undistort(image, mtx, dist):
image = cv2.undistort(image, mtx, dist, None, mtx)
return image
In [24]:
#Perspective transform
def transform(undist,src,dst,img_size):
M = cv2.getPerspectiveTransform(src, dst)
Minv = cv2.getPerspectiveTransform(dst, src)
warped = cv2.warpPerspective(undist, M, img_size)
return warped, M, Minv
In [25]:
def abs_sobel_thresh(image, orient, sobel_kernel, thresh):
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
assert(orient == 'x' or orient == 'y'), "Orientation must be x or y"
if orient == 'x':
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0,ksize = sobel_kernel)
abs_sobelx = np.absolute(sobelx)
scaled_sobel = np.uint8(255*abs_sobelx/np.max(abs_sobelx))
else:
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1,ksize = sobel_kernel)
abs_sobely = np.absolute(sobely)
scaled_sobel = np.uint8(255*abs_sobely/np.max(abs_sobely))
grad_binary = np.zeros_like(scaled_sobel)
grad_binary[(scaled_sobel >= thresh[0]) & (scaled_sobel <= thresh[1])] = 1
return grad_binary
def mag_thresh(image, sobel_kernel, thresh):
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize = sobel_kernel)
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize = sobel_kernel)
abs_sobelxy = np.power((np.power(sobelx,2)+np.power(sobely,2)),0.5)
scaled_sobel = np.uint8(255*abs_sobelxy/np.max(abs_sobelxy))
mag_binary = np.zeros_like(scaled_sobel)
mag_binary[(scaled_sobel >= thresh[0]) & (scaled_sobel <= thresh[1])] = 1
return mag_binary
def dir_threshold(image, sobel_kernel, thresh):
gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel)
with np.errstate(divide='ignore', invalid='ignore'):
absgraddir = np.absolute(np.arctan(sobely/sobelx))
dir_binary = np.zeros_like(absgraddir)
dir_binary[(absgraddir >= thresh[0]) & (absgraddir <= thresh[1])] = 1
return dir_binary
In [26]:
def get_thresholded_image(img):
# apply gradient threshold on the horizontal gradient
sx_binary = abs_sobel_thresh(img, 'x', 3, thresh=(10, 200))
# apply gradient direction threshold so that only edges closer to vertical are detected.
dir_binary = dir_threshold(img, 3,thresh=(np.pi/6, np.pi/2))
# combine the gradient and direction thresholds.
combined_condition = ((sx_binary == 1) & (dir_binary == 1))
# R & G thresholds so that yellow lanes are detected well.
color_threshold = 150
R = img[:,:,0]
G = img[:,:,1]
color_combined = np.zeros_like(R)
r_g_condition = (R > color_threshold) & (G > color_threshold)
# color channel thresholds
hls = cv2.cvtColor(img, cv2.COLOR_RGB2HLS)
S = hls[:,:,2]
L = hls[:,:,1]
# S channel performs well for detecting bright yellow and white lanes
s_thresh = (100, 255)
s_condition = (S > s_thresh[0]) & (S <= s_thresh[1])
# We put a threshold on the L channel to avoid pixels which have shadows and as a result darker.
l_thresh = (120, 255)
l_condition = (L > l_thresh[0]) & (L <= l_thresh[1])
# combine all the thresholds
# A pixel should either be a yellowish or whiteish
# And it should also have a gradient, as per our thresholds
color_combined[(r_g_condition & l_condition) & (s_condition | combined_condition)] = 1
# apply the region of interest mask
mask = np.zeros_like(color_combined)
region_of_interest_vertices = np.array([[0,img.shape[0]-1], [img.shape[1]/2, int(0.5*img.shape[0])], [img.shape[1]-1, img.shape[0]-1]], dtype=np.int32)
cv2.fillPoly(mask, [region_of_interest_vertices], 1)
thresholded = cv2.bitwise_and(color_combined, mask)
return thresholded
In [27]:
#Calibrating Camera
nx = 9
ny = 6
mtx, dist = camera_Calibraton('Users/rickerish_nah/Documents/trials/udacity Q/camera_cal', 'calibration', nx, ny, (720, 1280))
In [28]:
def get_averaged_line(previous_lines, new_line):
"""
This function computes an averaged lane line by averaging over previous good frames.
"""
# Number of frames to average over
num_frames = 12
if new_line is None:
# No line was detected
if len(previous_lines) == 0:
# If there are no previous lines, return None
return previous_lines, None
else:
# Else return the last line
return previous_lines, previous_lines[-1]
else:
if len(previous_lines) < num_frames:
# we need at least num_frames frames to average over
previous_lines.append(new_line)
return previous_lines, new_line
else:
# average over the last num_frames frames
previous_lines[0:num_frames-1] = previous_lines[1:]
previous_lines[num_frames-1] = new_line
new_line = np.zeros_like(new_line)
for i in range(num_frames):
new_line += previous_lines[i]
new_line /= num_frames
return previous_lines, new_line
def get_mean_distance_between_lines(left_line, right_line, running_average):
"""
Returns running weighted average of simple difference between left and right lines
"""
mean_distance = np.mean(right_line - left_line)
if running_average == 0:
running_average = mean_distance
else:
running_average = 0.9*running_average + 0.1*mean_distance
return running_average
In [29]:
class Line_line:
def __init__(self):
# Was the line found in the previous frame?
self.found = False
# Remember x and y values of lanes in previous frame
self.X = None
self.Y = None
# Store recent x intercepts for averaging across frames
self.x_int = deque(maxlen=10)
self.top = deque(maxlen=10)
# Remember previous x intercept to compare against current one
self.lastx_int = None
self.last_top = None
# Remember radius of curvature
self.radius = None
# Store recent polynomial coefficients for averaging across frames
self.fit0 = deque(maxlen=10)
self.fit1 = deque(maxlen=10)
self.fit2 = deque(maxlen=10)
self.fitx = None
self.pts = []
# Count the number of frames
self.count = 0
def found_search(self, x, y):
'''
This function is applied when the lane lines have been detected in the previous frame.
It uses a sliding window to search for lane pixels in close proximity (+/- 25 pixels in the x direction)
around the previous detected polynomial.
'''
xvals = []
yvals = []
if self.found == True:
i = 720
j = 630
while j >= 0:
yval = np.mean([i,j])
xval = (np.mean(self.fit0))*yval**2 + (np.mean(self.fit1))*yval + (np.mean(self.fit2))
x_idx = np.where((((xval - 25) < x)&(x < (xval + 25))&((y > j) & (y < i))))
x_window, y_window = x[x_idx], y[x_idx]
if np.sum(x_window) != 0:
np.append(xvals, x_window)
np.append(yvals, y_window)
i -= 90
j -= 90
if np.sum(xvals) == 0:
self.found = False # If no lane pixels were detected then perform blind search
return xvals, yvals, self.found
def blind_search(self, x, y, image):
'''
This function is applied in the first few frames and/or if the lane was not successfully detected
in the previous frame. It uses a slinding window approach to detect peaks in a histogram of the
binary thresholded image. Pixels in close proimity to the detected peaks are considered to belong
to the lane lines.
'''
xvals = []
yvals = []
if self.found == False:
i = 720
j = 630
while j >= 0:
histogram = np.sum(image[j:i,:], axis=0)
if self == Right:
peak = np.argmax(histogram[640:]) + 640
else:
peak = np.argmax(histogram[:640])
x_idx = np.where((((peak - 25) < x)&(x < (peak + 25))&((y > j) & (y < i))))
x_window, y_window = x[x_idx], y[x_idx]
if np.sum(x_window) != 0:
xvals.extend(x_window)
yvals.extend(y_window)
i -= 90
j -= 90
if np.sum(xvals) > 0:
self.found = True
else:
yvals = self.Y
xvals = self.X
return xvals, yvals, self.found
def radius_of_curvature(self, xvals, yvals):
ym_per_pix = 30./720 # meters per pixel in y dimension
xm_per_pix = 3.7/700 # meteres per pixel in x dimension
fit_cr = np.polyfit(yvals*ym_per_pix, xvals*xm_per_pix, 2)
curverad = ((1 + (2*fit_cr[0]*np.max(yvals) + fit_cr[1])**2)**1.5) \
/np.absolute(2*fit_cr[0])
return curverad
def sort_vals(self, xvals, yvals):
sorted_index = np.argsort(yvals)
sorted_yvals = yvals[sorted_index]
sorted_xvals = xvals[sorted_index]
return sorted_xvals, sorted_yvals
def get_intercepts(self, polynomial):
bottom = polynomial[0]*720**2 + polynomial[1]*720 + polynomial[2]
top = polynomial[0]*0**2 + polynomial[1]*0 + polynomial[2]
return bottom, top
In [30]:
# Video Processing Pipeline
def process_vid(image):
img = undistort(image,mtx, dist)
img_t=get_thresholded_image(img)
img_size = (img_t.shape[1], img_t.shape[0])
img_warp, M, Minv=transform(img_t, source, dst, img_size)
# Identify all non zero pixels in the image
x, y = np.nonzero(np.transpose(img_warp))
combined_binary = img_t.copy()
if Left.found == True: # Search for left lane pixels around previous polynomial
leftx, lefty, Left.found = Left.found_search(x, y)
if Right.found == True: # Search for right lane pixels around previous polynomial
rightx, righty, Right.found = Right.found_search(x, y)
if Right.found == False: # Perform blind search for right lane lines
rightx, righty, Right.found = Right.blind_search(x, y, img_warp)
if Left.found == False:# Perform blind search for left lane lines
leftx, lefty, Left.found = Left.blind_search(x, y, img_warp)
lefty = np.array(lefty).astype(np.float32)
leftx = np.array(leftx).astype(np.float32)
righty = np.array(righty).astype(np.float32)
rightx = np.array(rightx).astype(np.float32)
# Calculate left polynomial fit based on detected pixels
left_fit = np.polyfit(lefty, leftx, 2)
# Calculate intercepts to extend the polynomial to the top and bottom of warped image
leftx_int, left_top = Left.get_intercepts(left_fit)
# Average intercepts across n frames
Left.x_int.append(leftx_int)
Left.top.append(left_top)
leftx_int = np.mean(Left.x_int)
left_top = np.mean(Left.top)
Left.lastx_int = leftx_int
Left.last_top = left_top
# Add averaged intercepts to current x and y vals
leftx = np.append(leftx, leftx_int)
lefty = np.append(lefty, 720)
leftx = np.append(leftx, left_top)
lefty = np.append(lefty, 0)
# Sort detected pixels based on the yvals
leftx, lefty = Left.sort_vals(leftx, lefty)
Left.X = leftx
Left.Y = lefty
# Recalculate polynomial with intercepts and average across n frames
left_fit = np.polyfit(lefty, leftx, 2)
Left.fit0.append(left_fit[0])
Left.fit1.append(left_fit[1])
Left.fit2.append(left_fit[2])
left_fit = [np.mean(Left.fit0),
np.mean(Left.fit1),
np.mean(Left.fit2)]
# Fit polynomial to detected pixels
left_fitx = left_fit[0]*lefty**2 + left_fit[1]*lefty + left_fit[2]
Left.fitx = left_fitx
# Calculate right polynomial fit based on detected pixels
right_fit = np.polyfit(righty, rightx, 2)
# Calculate intercepts to extend the polynomial to the top and bottom of warped image
rightx_int, right_top = Right.get_intercepts(right_fit)
# Average intercepts across 5 frames
Right.x_int.append(rightx_int)
rightx_int = np.mean(Right.x_int)
Right.top.append(right_top)
right_top = np.mean(Right.top)
Right.lastx_int = rightx_int
Right.last_top = right_top
rightx = np.append(rightx, rightx_int)
righty = np.append(righty, 720)
rightx = np.append(rightx, right_top)
righty = np.append(righty, 0)
# Sort right lane pixels
rightx, righty = Right.sort_vals(rightx, righty)
Right.X = rightx
Right.Y = righty
# Recalculate polynomial with intercepts and average across n frames
right_fit = np.polyfit(righty, rightx, 2)
Right.fit0.append(right_fit[0])
Right.fit1.append(right_fit[1])
Right.fit2.append(right_fit[2])
right_fit = [np.mean(Right.fit0), np.mean(Right.fit1), np.mean(Right.fit2)]
# Fit polynomial to detected pixels
right_fitx = right_fit[0]*righty**2 + right_fit[1]*righty + right_fit[2]
Right.fitx = right_fitx
# Compute radius of curvature for each lane in meters
left_curverad = Left.radius_of_curvature(leftx, lefty)
right_curverad = Right.radius_of_curvature(rightx, righty)
# Only print the radius of curvature every 3 frames for improved readability
if Left.count % 3 == 0:
Left.radius = left_curverad
Right.radius = right_curverad
# Calculate the vehicle position relative to the center of the lane
position = (rightx_int+leftx_int)/2
distance_from_center = abs((640 - position)*3.7/700)
Minv = cv2.getPerspectiveTransform(dst, source)
warp_zero = np.zeros_like(img_warp).astype(np.uint8)
color_warp = np.dstack((warp_zero, warp_zero, warp_zero))
pts_left = np.array([np.flipud(np.transpose(np.vstack([Left.fitx, Left.Y])))])
pts_right = np.array([np.transpose(np.vstack([right_fitx, Right.Y]))])
pts = np.hstack((pts_left, pts_right))
cv2.polylines(color_warp, np.int_([pts]), isClosed=False, color=(0,0,255), thickness = 40)
cv2.fillPoly(color_warp, np.int_(pts), (34,255,34))
newwarp = cv2.warpPerspective(color_warp, Minv, (image.shape[1], image.shape[0]))
result = cv2.addWeighted(img, 1, newwarp, 0.5, 0)
# Print distance from center on video
if position > 640:
cv2.putText(result, 'Vehicle is {:.2f}m left of center'.format(distance_from_center), (100,80),
fontFace = 16, fontScale = 2, color=(255,255,255), thickness = 2)
else:
cv2.putText(result, 'Vehicle is {:.2f}m right of center'.format(distance_from_center), (100,80),
fontFace = 16, fontScale = 2, color=(255,255,255), thickness = 2)
# Print radius of curvature on video
cv2.putText(result, 'Radius of Curvature {}(m)'.format(int((Left.radius+Right.radius)/2)), (120,140),
fontFace = 16, fontScale = 2, color=(255,255,255), thickness = 2)
Left.count += 1
return result
In [31]:
Left=Line_line()
Right=Line_line()
white_output = "project_output.mp4"
clip1 = VideoFileClip('/Users/rickerish_nah/Documents/trials/udacity Q/project_video.mp4')
white_clip = clip1.fl_image(process_vid) #NOTE: this function expects color images!!
In [32]:
%time white_clip.write_videofile(white_output, audio=False)
In [33]:
Left=Line_line()
Right=Line_line()
chall_output = "challenge_output.mp4"
clip1 = VideoFileClip('/Users/rickerish_nah/Documents/trials/udacity Q/challenge_video.mp4')
chall_clip = clip1.fl_image(process_vid) #NOTE: this function expects color images!!
In [34]:
%time chall_clip.write_videofile(chall_output, audio=False)
In [35]:
HTML("""
<video width="960" height="540" controls>
<source src="{0}">
</video>
""".format(white_output))
Out[35]:
In [36]:
HTML("""
<video width="960" height="540" controls>
<source src="{0}">
</video>
""".format(chall_output))
Out[36]: